FR3053126B1 - METHOD AND DEVICE FOR LOCATING THE ORIGIN OF A DEFECT AFFECTING A STACK OF THIN LAYERS DEPOSITED ON A SUBSTRATE - Google Patents

Abstract

Method and device for locating the origin of a defect affecting a stack of thin layers deposited on a substrate The method according to the invention for locating in a deposition line comprising a succession of compartments, an origin of a defect affecting a stack of thin layers deposited on a substrate in the compartments, comprises: a step of obtaining (E10) at least one image representing said defect acquired by at least one optical control system placed at the exit of the line of deposit; a step of determining (E20), from said at least one image, a signature of the defect, this signature comprising at least one characteristic representative of the defect; an identification step (E40) of at least one compartment of the deposit line likely to be at the origin of the defect from the signature of the defect and using reference signatures associated with the compartments of the line deposit.

Description

BACKGROUND OF THE INVENTION The invention relates to the manufacture of substrates coated on at least one face with a stack of thin layers, in particular transparent substrates made of glass or of polymeric organic material.

It is conventional to provide substrates, in particular made of glass or of an organicpolymeric material, with coatings which confer on them particular properties, in particular optical properties, for example reflection or absorption of radiation of a given wavelength range, electrical conduction properties, or properties related to the ease of cleaning or the possibility for the substrate to self-clean. These coatings are generally stacks of thin layers based on inorganic compounds, especially metals, oxides, nitrides or carbides. For the purposes of the invention, a thin layer is a layer whose thickness is less than one micrometer and generally ranges from a few nanometers to a few hundred nanometers, hence the term "thin".

A stack of thin layers is generally manufactured via a succession of deposits of thin layers made in a plurality of compartments of a deposition line (typically 20 to 30 compartments), these deposits being made in the different compartments using one or more deposition methods. such as magnetic field sputtering (also known as magnetron sputtering), ion beam assisted deposition (IBAD), evaporation, chemical vapor deposition (or CVD for Chemical Vapor Deposition), plasma enhanced chemical vapor deposition (PECVD for Plasma-Enhanced CVD), chemical low pressure vapor deposition (LPCVD).

Unfortunately, the compartments of the deposit line are often dirty, and dust or debris present in some compartments may be caused to fall erratically onto the substrate as it passes through these compartments. Although an air blower mechanism is provided to remove such debris, some debris may occur on the surface of the substrate (more precisely on the surface of the thin layer deposited in the compartment in question) and then act as masks for the deposits. subsequent thin layers. These debris are at the origin of defects affecting the quality of the stack of thin layers deposited on the substrate and which can be prohibitive depending on the intended application for the coated substrate thus manufactured.

In order to control the quality of the coated substrate manufactured, there exist in the current state of the art optical control systems intended to be placed at the output of the deposition line, and which are configured to provide different images of the thin film stack. deposited on the surface of the substrate. These optical control systems are generally equipped with a bench comprising a plurality of optical sensors (cameras) and several radiation sources at different wavelengths, allowing the acquisition of images in differentconfigurations (eg images in reflection, in transmission, etc.). The analysis of these images makes it possible to detect any defects affecting the stack of thin layers made in the deposition line.

Although the control systems proposed today allow the detection of faults, they do not give any information on the origin of these faults. However, knowledge of this origin can prove invaluable to consider targeted and rapid maintenance operations on the deposit line.

OBJECT AND SUMMARY OF THE INVENTION The invention makes it possible in particular to overcome the drawbacks of the state of the art by proposing a method of locating, in a filing line comprising a succession of compartments, of an origin of a defect affecting a stack. thin film deposited on a substrate in the compartments, the method comprising: - a step of obtaining at least one image representing said defect acquired by at least one optical control system placed at the output of the deposition line; A step of determining, from said at least one image, a signature of the defect, this signature comprising at least one representative characteristic of the defect; A step of identifying at least one compartment of the deposit line likely to be at the origin of the defect from the signature of the defect and using reference signatures associated with the compartments of the deposit line.

Correlatively, the invention also provides a device for locating a fault origin affecting a stack of thin layers deposited on a substrate in a plurality of compartments succeeding one another in a deposition line, this device comprising: a module for obtaining at least one image representing the defect acquired by at least one optical control system placed at the output of the deposition line; A module for determining, from said at least one image, a signature of the defect, this signature comprising at least one characteristic representative of the defect; An identification module of at least one compartment of the deposit line likely to be at the origin of the defect from the signature of the defect and using signatures of reference associated with the compartments of the deposit line. The invention thus proposes a simple and effective solution for locating the origin in a deposition line of a defect affecting a stack of thin layers deposited on a substrate.

The substrate is preferably a mineral glass sheet or an organic polymeric material. It is preferably transparent, colorless or colored. The glass is preferably of the silico-soda-lime type, but it may also be, for example, a borosilicate or alumino-borosilicate type glass. The preferred polymeric organic materials are polycarbonate, polymethyl methacrylate, polyethylene terephthalate (PET), polyethylene naphthalate (PEN), or fluorinated polymers such as ethylene tetrafluoroethylene (ETFE). The substrate may be rigid or flexible. The substrate may be flat or curved.

The solution according to the invention is based on the exploitation of digital images resulting from an optical control conventionally performed at the output of the deposition line. According to the invention, these images are analyzed in order to extract a signature of the defect comprising one or more predetermined characteristics, representative of the defect, and the observation of which makes it possible to identify compartments of the deposit line likely to be the origin of the defect. For this purpose, the signature of the defect is approximated directly or indirectly (by way of example of one or more decision trees) of different reference signatures, generated for each of the compartments of the deposit line and containing the same characteristics as the default signature. . A reference signature associated with a compartment represents the predetermined characteristic values of a default originating in that compartment. It can be obtained for example experimentally from a noise made on the walls of the compartment, by simulation or by calculation. It is noted that several reference signatures can be associated with each of the compartments of the deposit line. Likewise, each reference signature does not necessarily correspond to a single point but may correspond to a range of values, a portion of a curve, a portion of a surface, and so on. following the number of components included in the reference signature. The analysis of the signature of the defect, taking into account the reference signatures associated with the different compartments of the deposit line, makes it possible to identify a small number of compartments of the deposit line, or even a single compartment, likely to be at the origin of the deposit. default.

Thus, for example, in a particular embodiment, the identification stepcomprises a step of comparing the signature of the defect with a plurality of reference signatures associated with each compartment of the deposit line, said at least one compartment identified as being which may be at the origin of the defect being associated with a signature of reference corresponding to the signature of the defect. In other words, the compartment or compartments identified as being likely to be at the origin of the defect are in this case the compartments whose reference connections are closest to the signature of the defect, in the sense, for example, of a predefined distance or so equivalent, a predefined prediction error. The invention advantageously applies to different stacks of thin layers that can be deposited on a substrate, in particular a transparent substrate, and in particular to a stack of thin layers forming an interference system. The choice of one or more characteristics of the signature of the defect used in the context of the invention is adapted depending on the optical properties of the thin-film stack considered, in particular the reflection or radiation transmission properties, as well as the type of images provided by the optical control system (images in reflection, in transmission, obtained with radiation sources that can emit in different wavelength ranges, etc.). Examples of stacks of thin layers that can be deposited on a substrate and analyzed in accordance with the invention include, but are not limited to: - Stacks of thin layers that modify the reflection properties of the substrate in the area visible wavelengths, such as reflective metallic layers, especially those based on metallic silver, which are used to form the mirrors, or anti-reflective coatings, which aim to reduce the radiation reflection at the interface between air and the substrate. An antireflection coating may be formed, in particular, by a stack of thin layers having refractive indices alternatively lower and stronger playing the role of an interference filter at the interface between the air and the substrate, or by a stack of layers thin films having a gradient, either continuous or staggered, of indices of refraction between the refractive index of air and that of the substrate; Stacks of thin layers which give the substrate infrared radiation reflection properties, such as transparent stacks, comprising at least one thin metal layer or electroconductive transparent oxide (TCO) layer, called a functional layer, in particular based on silver, niobium, chromium, nickel-chromium alloy (NiCr), indium mixed oxide and tin (ΓΓΟ), and coatings located on either side of each functional layer to form a system interference. Thesetransparent stacks with infrared reflection properties are used to form solar control glazing, especially anti-solar, aimed at reducing the amount of incoming solar energy, or low emissivity, aimed at reducing the amount of energy dissipated to the outside a building or a vehicle; Stacks of thin layers which confer on the substrate electrical conduction properties, such as transparent stacks, comprising at least one metallic thin layer, in particular based on silver, or a thin layer based on transparent electroconductive oxides (TCO), for example based on mixed tin and indium oxide (ΓΓΟ), based on mixed indium zinc oxide (IZO), based on zinc oxide doped with gallium or aluminum, with niobium doped titanium oxide base, based on cadmium stannate or dezinc, based on fluorine doped tin oxide and / or antimony. These coatings with electrical conduction properties are used, in particular, in heated glazing units, where an electric current is circulated in the coating so as to generate heat Joule effect, or as an electrode in electronic devices to layers, in particular as a transparent electrode located on the front face of organic light-emitting diode (OLED) devices, photovoltaic devices, electrochromic devices; Stacks of thin layers which give the substrate self-cleaning properties, such as transparent titanium oxide stacks, which facilitate the degradation of organic compounds under the action of ultraviolet radiation and the elimination of mineral soiling under the action of a runoff of water; Stacks of thin layers which confer on the substrate properties of anticondensation, hydrophobic properties, etc.

The characteristic (s) chosen to form the signature of the defect preferably has a value which varies according to the compartments participating in the deposition of the stack of thin layers, so as to easily allow the identification of a compartment at the origin of a defect in this stack. For example, a characteristic whose value is a strictly monotonic (ie, increasing or decreasing) function of the thickness of the deposit on the substrate can be chosen, so that the value of this characteristic in the signature of the defect can identify a deposit thickness to which this one appeared. From the knowledge of this thickness and the compartment in charge of depositing a thin layer corresponding to this thickness, it is easy to deduce the compartment at the origin of the defect. This choice of a strictly monotonic characteristic as a function of the thickness of deposits on the substrate is particularly advantageous because it allows from a signature comprising a single well-chosen characteristic to identify a single compartment at the origin of the defect.

An unequivocal relationship between the value of each characteristic of the signature and a compartment of the deposit line, however, is not an essential condition of the invention, that is, it is not necessary for each compartment to correspond to a distinct value of each feature. An ambiguity, or degeneracy, between two or more compartments can indeed subsist when considering a given characteristic. In particular, the same value of a characteristic may correspond to two separate compartments of the deposit line.

Indeed, it is noted first of all that, even in the case where there remains an ambiguity, the invention makes it possible to drastically reduce the number of compartments suspected of being at the origin of the defect. As mentioned above, a deposition line in the case of unpacking a stack of thin layers on a substrate can easily comprise 20 or even 30 compartments. The invention allows to be limited to a few compartments for which an ambiguity remains, that is to say, by choosing the characteristics, to at most two or three compartments. It is also possible to further reduce this number by considering several features in the signature of the defect, the multiplicity of features to remove residual ambiguities.

It is noted that the characteristics considered to form the signature of a defect may depend on the nature of the images obtained and provided by the optical systems (eg images in reflection or in transmission, types of radiation sources used to generate the images, etc.) . The use of images of different natures can advantageously allow to overcome the aforementioned ambiguities by extracting from each type of images different additional characteristics to form the signature of the defect.

Said at least one image representing the defect can be, in particular, a grayscale image, an RGB encoded image, a hyperspectral image.

Thus, by way of illustration, in a particular embodiment, said at least one image may comprise a gray-scale coded reflection image and a grayscale-coded transmission image.

This embodiment may have a preferred but nonlimiting application when the thin-film stack comprises several layers having a low emissivity. The use of information extracted from a transmission image in addition to the information extracted from the image in reflection allows to have a signature comprising more characteristics, and to be able to more precisely locate the origin of the defect affecting the stack of thin layers.

In another embodiment, said at least one image may comprise two grayscale-encoded images (in reflection and / or transmission) acquired by said at least one optical system using two radiation sources emitting in two wavelength less partially distinct, including a radiation source emitting in the visible wavelength range and a radiation source emitting in the wavelength range of the infrared. The combination of images obtained with the help of two sources of radiation emitting at different wavelengths, which are chosen with great care, can make it possible to remove an ambiguity.

Some stacks may indeed have strictly monotone characteristics as a function of the thickness of deposition in the infrared range that this embodiment advantageously allows to exploit. The invention therefore makes it possible to carry out targeted maintenance on the deposit line, enconcentrating the maintenance operations (and in particular cleaning operations) or, more generally, the corrective actions on the few identified compartments, or even a single compartment. Various corrective actions may typically consist of knocking on the walls of each identified compartment in order to rid them of any debris and dust that may collect them and which may deposit on the substrate during the deposition of the mincerealized layer in this compartment. These maintenance operations can take place indifferently during production, or at the end of production. The invention thus makes it possible to intervene more quickly and more efficiently on the deposit line, while limiting the associated costs. The identification of the compartment or compartments causing defects is also very fast thanks to the invention. This results in a significant time saving in the resolution of crisis when detecting the appearance of defects on the coated substrates from the deposition line.

In a particular embodiment: said at least one image comprises a reflection image in grayscale, and the fault signal includes a characteristic defined from a reflection coefficient of the defect determined from the image in reflection ; and / or - said at least one image comprises a grayscale encoded transmission image, andthe signature of the defect comprises a characteristic defined from a transmission coefficient of the defect determined from the transmission image.

These reflection and / or transmission coefficients can be determined, for example, by using a successive erosion method applied to the defect represented on the grayscale-coded image, during which, at each step, a reflection coefficient value and / or transmission is determined. Other methods may well be envisaged alternatively.

The inventors have judiciously discovered that it is possible, from the reflection image or the grayscale-encoded transmission image of the defect and more particularly from a level of contrast in reflection or transmission of the determined defect. from this image, to identify which compartment (s) could be the result of this defect. More particularly, the inventors have established a correlation between the level of contrast in reflection or transmission of the defect and the thickness at which the stack was stopped following the fall of a dust or debris, in other words the thickness at which appeared the defect. This thickness can be easily associated with a compartment of the deposition line participating in deposition of the stack of thin layers. Note that when the image is not hyperspectral, the invention is based on the assumption that the contrast value considered extracted from the images provided by the optical control system reflects the "true" integrated value of contrast (reflection or transmission ) which would be calculated from the spectrum (of reflection or transmission). By "translated" it is meant that the contrast value considered is representative of this true value, either absolutely or relative (i.e. it may be proportional to this true value or be a monotonic function thereof). As an illustration, in the case of a transparent stack of thin films with infrared reflection properties as described above, comprising a thin layer of silver framed by two coatings each comprising one or more thin layers to form an interference system, it is possible to observe an evolution of the average light reflection coefficient observed on the reflection image (possibly normalized) as a function of the thickness of the stack having the shape of an inverted V (A) and presenting a maximum at the level of the thin layer of money. This maximum at the level of silver layer is explained by the low emissivity of the silver layer, which results in maximum light reflection. In other words, the evolution of the light reflection coefficient as a function of the thickness is an increasing function before the silver layer and decreasing thereafter. It is therefore easy to discriminate, from this representation and the knowledge of so-called reference light reflection coefficients characterizing each of the compartments of the deposit line, a compartment likely to be at the origin of the defect parmiles compartments of a first a series of compartments participating in the deposit of the first deposit deposited before the silver layer, or a compartment likely to be at the origin of the defect among the compartments of a second series of compartments participating in the deposit of the second coating deposited after the coating of money.

Thus, in the example illustrated above, if the signature of the defect comprises a unique characteristic defined from the average light reflection coefficient of the defect, two compartments are obtained at the end of the process according to the invention which can be the origin of the defect found on the stack of thin layers.

In order to determine whether the compartment causing the observed defect belongs to the first or second series of compartments, the inventors advantageously propose to exploit a second characteristic of the defect, namely, in the example mentioned above, the presence or absence of a ring. clear (ie corresponding to a maximum reflection) surrounding the defect on the reflection image.

In a particular embodiment of the invention, the locating method also comprises a step of detecting a presence of a light ring on the excessive defect represented on said at least one image, the signature of the defect comprising a transliterating characteristic. this presence.

The inventors have indeed connected the presence of such a clear ring with the introduction of a defect before the deposition of the silver layer, ie in a compartment of the first series of compartments participating in the deposition of the first coating deposited before the layer. of silver, the absence of such a clear ring on the contrary translating the introduction of a defect after the deposition of the silver layer, ie in a compartment of the second series of compartments taking part in the depositing of the second coating deposited after the layer of money. It is noted that the presence of a clear ring can easily be detected through the use of the erosion method mentioned above, by observing the values of the light reflection coefficient obtained at the different iterations of the erosion method, and in particular during first iterations.

This also applies when the optical system provides an image transmitting the defect and a characteristic of the signature of the defect is determined from a light transmission coefficient of the defect.

In a particular embodiment of the invention, the method further comprises: a step of determining a coefficient variation gradient (reflection or transmission); and a step of detecting a form of the defect from the determined variation gradient.

This embodiment makes it possible to obtain additional information on the defect, especially if it is a flat defect or a three-dimensional defect. Such information may be useful for identifying, within a compartment, which elements of the compartment are responsible for the dust and / or debris deposited on the substrate. The invention applies, as mentioned above, to images coded at a level of degree (eg images in reflection and / or in transmission, acquired by means of a radiation source emitting in the wavelength domain of the visible and / or in the wavelength range of the infrared, etc.). However, this hypothesis is not limiting and the invention can also be implemented from other types of images. The signatures considered are then adapted to the information that can be extracted from these images.

Thus, in a particular embodiment, said at least one image comprises an image in reflection or transmission coded in red, green and blue (RGB), and the method further comprises a step of converting the RGB image into a colorimetric space. * a * b *, the default signature comprising components at a * and b * of a defect background surface determined from the converted image. The signature of the defect may also comprise a component in L * of the defect determined from the converted image.

In another embodiment, said at least one image comprises a hyperspectral image and the signature of the defect comprises a spectrum representing values of a coefficient of reflection or of transmitting a background surface of the defect as a function of a wavelength. . Moreover, each reference signature associated with a compartment of the deposit line may comprise, in this embodiment, a plurality of spectra corresponding to different thicknesses of the layer deposited in the compartment.

These different types of images, encoded in RGB or hyperspectral, provide more defect information than grayscale encoded images. However, they require the use of more complex and often expensive optical control systems. The invention makes it possible to adapt to different optical control systems placed outside the deposit line.

In the aforementioned embodiments, a reduced number of characteristics representative of the defect has been extracted from the images of this defect and exploited. This reduced number of features limits the complexity of the identification step, since only a few features are to be compared to the reference signatures of the compartments during this step.

In another embodiment of the invention, the identification step comprises an application of an automatic learning method on the signature of the defect, said learning method being based on a model driven from the reference signatures. associated with the compartments of the deposit line. The training of the model is preferably performed on a large number of imagesrepresenting defects whose origin we are able to identify or whose origin is known, these images having been obtained over several days of production. Such a model is, for example, composed of one or more decisional trees driven from reference signatures of the compartments extracted from the large number of images. The automatic learning method used during the identification step can be for example a so-called decision tree forest algorithm also known as the "random decision forest" in English able to use such decision trees. However, any other algorithm for classifying elements between them and based on a model driven from reference values can be used alternatively (eg nearest neighbor algorithm, support vector machines or Support Vector Machine (SVM) in English, neural networks, etc.).

In this embodiment, outstanding characteristics of the defect other than those mentioned above can be easily extracted from the digital images and exploited to identify the compartment at the origin of the defect. Preferably, the signature of the defect determined from said at least one image comprises at least one characteristic of the light intensity of the defect and / or a characteristic relating to a form of the defect. By way of example, said at least one light intensity characteristic of the defect can comprise: characteristics representative of a radial profile of light intensity of the defect; and / or - characteristics representative of a slope of a radial profile of light intensity of the defect; and / or - a characteristic representative of a mean light intensity of the defect; and / or - a characteristic representative of a luminous intensity at the center of the defect.

Similarly, said at least one characteristic relating to a form of the defect may comprise: a characteristic representative of a defect area; and / or - a characteristic representative of a ratio of a defect perimeter to a defect area, and / or - a characteristic representative of a defect form factor.

The aforementioned features have been identified by the inventors as allowing, when used in whole or in part, to very reliably identify the compartment at the origin of the defect. It should be noted that the use of an automatic learning method makes it easy to take into account a larger number of characteristics of the defect.

It should be noted that in all the embodiments described above of the invention, the correspondences between the reference signatures and the compartments are likely to evolve, and depend in particular on the setting of the deposition line, during the deposition of the thin layers on the substrate. Indeed, the contribution of each cathode to the thickness of the stack can be made to be modified, in particular due to changes in the power applied on the cathode, the speed of travel of the substrate with respect to the cathode, the pressure of the gas used in the cathode compartment and its composition, etc.This dependence on the deposition parameters can be easily taken into account empirically or analytically, in particular by means of a calculation model linking the parameters of the deposit at the thickness of the stack corresponding to each compartment.

In a particular embodiment, the different steps of the delocalization process are determined by computer program instructions.

Consequently, the invention also relates to a computer program on a recording medium or information medium, this program being capable of being implemented in a localization device or more generally in a computer, this program including instructions adapted to the implementation of the steps of a location method as described above.

This program can use any programming language, and be in the form of source code, object code, or intermediate code between source code and object code, such as in a partially compiled form, or in any other desirable form . The invention also relates to a computer readable information or recording medium, and comprising instructions of a computer program as mentioned above.

The information or recording medium may be any entity or device capable of storing the program. For example, the medium may comprise a destocking means, such as a ROM, for example a CD ROM or a microelectronic circuit ROM, or still a magnetic recording means, for example a floppy disk or hard disk. On the other hand, the information or recording medium may be a transmissible medium such as an electrical or optical signal, which may be conveyed via an electrical or optical cable, by radio or by other means. The program according to the invention can in particular be downloaded on an Internet type network.

Alternatively, the information or recording medium may be an integrated circuit in which the program is incorporated, the circuit being adapted to execute or to be used in the execution of the method in question. The invention also relates to a system comprising: - a deposition line comprising a succession of compartments able to deposit a stack of thin layers on a substrate; At least one optical control system placed at the output of the deposition line configured toprovide at least one image representing a defect affecting the stack of thin layers deposited on the substrate; and a locating device according to the invention, capable of identifying among the succession of compartments of the deposit line at least one compartment likely to be at the origin of the defect.

It may also be envisaged, in other embodiments, that the delocalization process, the locating device and the system according to the invention present in combination all or part of the aforementioned characteristics.

BRIEF DESCRIPTION OF THE DRAWINGS Other features and advantages of the present invention will emerge from the description given below, with reference to the accompanying drawings which illustrate an embodiment thereof without any limiting character. In the figures: FIG. 1 schematically represents a system according to the invention, in a particular embodiment; FIG. 2 represents a deposition line of a stack of thin layers included in the system of FIG. 1; FIG. 3 represents an optical control system placed at the output of the deposition line in the system of FIG. 1; FIG. 4 schematically represents the hardware architecture of a delocalization device according to the invention included in the system of FIG. 1, in a particular embodiment; FIGS. 5, 11, 14 and 16 represent, in the form of flow charts, four embodiments of the localization method according to the invention; FIG. 6 represents a successive erosion method that can be applied to an image of a defect affecting a stack of thin layers to determine the cedefault signature in a particular embodiment; FIGS. 7, 9A-9C, 10, 12, 13, and 15 show various examples of reference signatures that can be used to identify the compartment causing a fault affecting a thin-film stack in accordance with the invention; and FIG. 8 illustrates a compartment comprising a cathode, as well as the various elements protecting this cathode.

Detailed description of the invention

FIG. 1 represents, in its environment, a system 1 according to the invention, in a particular embodiment.

The system 1 comprises: - a deposition line 2 for depositing a stack of thin layers on a transparent substrate 3; An optical control system 4 placed at the output of the deposition line 2, and able to acquire various digital images of the transparent substrate coated with the stack of thin layers (referenced by 5), and to detect from these images the presence of defects affecting the coated substrate 5; and a locating device 6 according to the invention, capable of analyzing the digital images IM of the defects detected by the optical control system 4 in order to locate their origin in the deposition line 2.

In the examples envisaged hereinafter, the stack of thin layers is deposited on a glass substrate 3 and forms an interference system. In these examples, the thin-film stack comprises one or more functional layers with infrared reflection properties (ie one or two layers of silver in Examples 1 and 2, a layer of oxides of indium and tin (ΓΓΟ) in Example 3), and coatings formed of one or more thin layers located on either side of each functional layer to form the interfering system. In the following, the term "module" (Ml, M2, M3) denotes each of the coatings which surround the functional layer (s) of silver or dTTO, it being understood that a module may consist of a single thin layer or d a plurality of thin layers.

Of course, the examples described hereinafter are not limiting in themselves and other stacks of thin layers as well as other substrates, especially transparent substrates (ex.substrats of polymeric organic material, flexible or rigid), can be envisaged.

In the examples below, the coatings or "modules" which frame the functional layer or layers consist of thin layers based on dielectric materials (eg thin layers of oxide, nitride, oxynitride). As a variant, especially in the case of metal thin layers as functional layers, each module which is disposed above, in the depositing direction of the stack, with a thin metallic layer may comprise, as an overlying layer to the thin metallic layer, a thin metal layer of over-blocker, oxidized or not, for protecting the thin metal layer during the deposition of a cocheultrière, for example if the latter is deposited under an oxidizing or nitriding atmosphere, and during a possible subsequent heat treatment. Each thin metal layer may also be deposited on and in contact with a thin sub-blocker metal layer. The thin-film stack may therefore comprise an over-blocking layer and / or a sub-blocking layer flanking the or each thin layer. metallic. These layers of deblocker, which are very thin layers, normally of a thickness less than 1 nm not to affect the light transmission of the stack, act as sacrificial layers, especially capable of capturing oxygen.

In the embodiment illustrated in FIG. 1, the deposition of thin layers is carried out by the deposition line 2 on the glass substrate 3 by means of a magnetic field assisted sputtering technique also known as magnetron sputtering.

In known manner, a sputtering technique relies on the condensation within a rarefied atmosphere of a vapor of a target material from a source of sputtering on a substrate. More precisely, the atoms of the source (also referred to as a target) are ejected in an ionized gas, such as, for example, argon, into a sousvide enclosure maintained at a certain pressure. An electric field is created leading to the ionization of gas thus forming a plasma. The target is brought to a negative potential (cathode) so that lesions present in the plasma are attracted to the target and eject atoms from it. The particles thus sprayed are diffused in the chamber and some of them are collected in particular on the substrate on which they form a thin layer. In the case of a cathodemagnetron, a magnetic field oriented perpendicular to the electric field is also created by magnets placed near the cathode so as to confine the electrons in the vicinity of the cathode. This makes it possible to increase the ionization rate of the gas, and thus to significantly improve the deposition efficiency compared to a conventional sputtering technique. Since sputtering techniques are known to those skilled in the art, they are not described in more detail here.

FIG. 2 diagrammatically represents the deposition line 2 used for producing the thin film deposit by cathode sputtering on the glass substrate 3. It comprises an inlet chamber 7, a first buffer chamber 8, a magnetron sputtering chamber 10 comprising a first transfer section 9 and a second transfer section 11, a second buffer chamber 12 and an exit chamber 13.

The sputtering chamber 10 comprises, in addition to the two transfer sections 9 and 11, a succession of elements Ei, i = 1,..., N where N denotes an integer. Each element Ei comprises a compartment or deposition chamber 15-i containing a cathode used as target during magnetron sputtering, and one or two compartment (s) or chamber (s) pump (s) equipped (s) with a pump, and located (s) where appropriate on both sides of the deposition chamber to create the vacuum therein. The glass substrate 3 circulates in the different successive compartments of the sputtering chamber 10, driven by a conveyor or a conveyor belt 16.

The succession of the different elements Ei, i = 1,..., N makes it possible to deposit a stack of thin layers on the glass substrate 3. It should be noted that several 15-isuccessive compartments can participate in the deposition of a layer thin of the same material, in predefined proportions. The parameterization of the various compartments 15-i in which the thin-film deposits take place consists in adjusting various parameters of the magnetron cathode sputtering, and in particular, the pressure of the gas and its composition, the power applied to the cathode, the angle of incidence of the bombardment particles, the thickness of the deposit, etc.

The optical control system 4 is placed at the output of the deposition line 2. As schematically illustrated in FIG. 3, it is equipped with one or more cameras 17 and with several radiation sources 18 making it possible to acquire and generate different types of radiation. digital images of the coated substrate 5 coming from the deposition line 2. In the example envisaged in FIG. 3, the optical control system 4 comprises three radiation sources 18, namely, a light source 18-1 of the RDF type (FIG. for Dark Field Reflection or Dark Field Reflection), an 18-2 light source of the RBF (Reflection Bright Field) type, and an 18-3 light source of TBF type (for Transmission Bright Field or Clear Encase Transmission). ). These light sources are switched on and off alternately in order to acquire different configurations of digital images in reflection (corresponding to the reflection of the light source on the coated substrate) and / or in transmission (corresponding to the light from the light source and transmitted through the coated substrate) representing the potential defects affecting the coated substrate comprising the thin film stack.

No limitation is attached to the nature of the cameras 17, resulting digital images, and the configuration of the radiation sources. Thus, according to their nature and their capabilities, the cameras 17 can provide, for example, digital images coded in gray levels, trichromatic images coded in RGB (Red Green Blue), or it can still be hyperspectral cameras capable of providing images. hyperspectrals, etc. On the other hand, radiation sources operating in the visible or other wavelength range, such as in the infrared field, can be used by the optical control system 4. These radiation sources can be oriented according to different angles according to the images that one wishes to acquire and use to locate the compartment or compartments at the origin of the defect.

FIG. 6 shows, in the lower part, by way of illustration, a digital reflection image, coded in gray levels, acquired from an RBF light source operating in the visible field. The task appearing on this image reflects a defect detected by the optical control system 4. In this image, the contrast measured by the camera of the optical control system 4 is proportional to the light reflection of the defect. It should be noted that when the image is not hyperspectral as is the case here, the invention is based on the assumption that the contrast value considered extracted from the images provided by the optical control system reflects the value of the integrated quantity ( RL or TL) that would be extracted from a spectral response (either relative or absolute).

Such an optical control system is known per se and conventionally used to detect defects affecting the deposits made in a deposit line such as the debit line 2. It is therefore not described in more detail here. It is noted that the invention is not limited to the use of a single optical control system and it is conceivable that the images processed by the localization device according to the invention come from a plurality of optical control systems placed at the output of the deposit line.

As mentioned above, the defects affecting the coated substrate may be due in particular to dusts or debris present in some compartments 15 of the sputtering chamber 10 which have fallen erratically onto the substrate as it has been circulating in these compartments. compartments. Very advantageously, by exploiting the images of the defects detected on the coated substrates 5 coming from the deposition line 2, the locating device 6 according to the invention is capable of locating the compartment 15-i of the deposition line 2 to the origin of this defect.

In the embodiment described here, the location device 6 is a computer whose hardware architecture is shown schematically in FIG. 4. It comprises a processor 19, a random access memory 20, a read-only memory 21, a non-volatile memory 22, a module 23, and various input / output modules 24.

The communication module 23 enables the localization device 6 to obtain the images of faults acquired by the optical control system 4. It can comprise in particular a digital data bus, and / or means of communication on a network (local or remote). such as, for example, a network card, etc., depending on how the optical control system 4 and the location device 6 are connected to each other.

The input / output modules 24 of the locating device 6 comprise in particular a keyboard, a mouse, a screen, and / or any other means (eg a graphical interface) making it possible to configure the locating device 6 and to access the results of the analysis that leads to the images of defects that are provided.

The read-only memory 21 of the locating device 6 constitutes a recording medium in accordance with the invention, readable by the processor 19 and on which is recorded a computer program PROG according to the invention, comprising instructions for the execution of the steps of FIG. a localization method according to the invention.

This computer program PROG equivalently defines functional and software modules here configured to implement the steps of the localisations methodaccording to the invention. These functional modules support or control the hardware elements 19 to 24 cited above. They include in particular: - a obtaining module 6A of at least one image representing a defect, acquired by the optical control system 4, this module 6A being able to communicate and control the communication module 23; A determination module 6B, from said at least one image, of a fault signature comprising at least one characteristic representative of the defect; and an identification module 6C of at least one compartment likely to be at the origin of the defect configured to use the signature of the defect and a plurality of reference signatures associated with the compartments of the deposit line.

The functions of these modules are described in more detail later in reference numerals 5 to 16, in four separate embodiments of the invention. The first three embodiments are distinguished by the nature of the digital images that are provided by the optical control system 4 and analyzed by the location device 6 to locate the origin of a fault. Thus, in the first embodiment, the images used by the delocalization device 6 are digital images coded in gray level; in the second embodiment, these are trichromatic digital images (or RGB for Red Green Blue), and in the third embodiment, the images provided by the optical control system are hyperspectral images. In the fourth embodiment, the locating device 6 implements an alternative treatment with respect to the first three modes for locating the defect in the deposition line 2. These four embodiments of the invention, however, rely on the same main steps to locate the defect, namely: a step of obtaining one or more images representing said defect and acquired by the optical control system 4; A step of determining, from these images, a signature of the defect comprising at least one representative characteristic of the defect; and a step of identifying at least one compartment among the compartments 15-i, i = 1,..., N, of the deposit line likely to be at the origin of the defect from the signature of the defect and using the reference signatures associated with compartments 15-i, i = 1, ..., N.

We will now describe in more detail these steps in each of the embodiments. To illustrate these different modes, the following three examples of thin film stacks are considered alternately. - Example 1: the stack is deposited on a soda-lime-silica glass substrate andcomprises, in the depositing direction of the stack on the substrate: a first coating or module M1 formed by a plurality of thin layers based on dielectric materials (eg oxide, nitride, oxynitride layers); o a thin layer of silver; and o a second coating or module M2 formed by a plurality of thin layers based on dielectric materials (eg oxide, nitride, oxynitride layers); each layer of the stack being able to be deposited by several different cathodes in order to obtain the required thickness, and the geometrical thicknesses of the thin layers of the stack being adapted so that the whole of the stack forms an interferential system. - Example 2: the stack is deposited on a soda-lime-silica glass substrate and comprises two thin layers of silver framed by thin layers of Si3N4, with the following thicknesses:

The geometrical thicknesses of the thin layers of the stack are adapted such that the entire stack forms an interference system. In this example 2, different thin layers of the same material (Si3N4 or Ag) are deposited on the substrate at different thicknesses. The layer to which reference is made is indicated in parentheses, for example Ag (Ll) denotes the first layer of silver deposited and Ag (L2) the second layer of silver. For example 2, "module M1" is also referred to as the Si3N4 thin film deposited on the substrate before the first silver layer Ag (L1), the "M2 module" being the thin layer of Si3N4 deposited between the first silver layer. Ag (Ll) and the second

silver layer Ag (L2), and "modulus M3" the thin layer of Si3N4 deposited after the second layer of silver Ag (L2). - Example 3: the stack is deposited on a soda-lime-silica glass substrate and comprises a thin layer of indium mixed oxide and tin (ΓΤΟ) framed by coatings ormodules M1, M2 based on dielectric materials , with the following thicknesses:

The geometrical thicknesses of the thin layers of the stack are adapted such that the entire stack forms an interference system. In this example 3, different thin layers of the same material are deposited on the substrate at different thicknesses. The layer to which reference is made in parentheses. For example 3, "module M1" is called the coating formed by the SiN and SiO2 thin layers (L1) deposited on the substrate before the couche layer, and "M2 module" the coating formed by the SiO2 thin layers. (L2) and TiO2 deposited on the substrate after the de layer.

Of course these examples are given for illustrative purposes only and are in no way limitative of the invention. As mentioned above, the invention applies to various stacks of thin layers that can be deposited on a substrate, and in particular to the stacks forming an interference system.

It is assumed in the embodiments described below that a defect has been detected by means of the images acquired by the optical control system 4. The manner in which the defect is detected is known per se and is not described here.

FIG. 5 represents the main steps of the localization method according to the invention as it is implemented by the localization device 6 in the first embodiment.

In this first embodiment of the invention, the optical control system 4is equipped with digital cameras capable of providing gray-level IM digital images of the detected defect, acquired by activating all or part of the various sources of radiation of the control system. 4. Such images consist of a plurality of pixels, each pixel being associated with a gray level reflecting its brightness.

We are interested in IMR defect-based digital images acquired by activating the RBF light source 18-2 and reflecting the reflection coefficient.

of the defect, as well as to digital images in transmission of the defect designated by IMTacquises by activating the light source 18-3 of type TBF and representing a transmission coefficient of the defect. For the sake of simplification, in the remainder of the description, the following is limited to an image in IMR reflection and, if appropriate, to an image in IMT transmission.

Note that as further detailed below, depending on the thin layer stacking configurations considered, it is not necessary for the location device 6 to systematically arrange both types of images.

In the case of a stack similar to Example 1 comprising a single layer of silver (or of another low-emissivity material), a single image in IMR reflection can suffice indeed for a precise location of the origin of the defect. .

In Example 2 comprising several silver layers, the use of a single image in IMR reflection can isolate a small number of compartments likely to be the cause of a defect but is not always sufficient to help identify a unique compartment among this reduced number of compartments. To eliminate this ambiguity, another image than a reflection image can be used, such as a transmission image acquired by means of a radiation source operating in the infrared wavelength domain, or it can be used several images of the defect such as for example an image in IMR reflection and an image in IMT transmission both acquired with radiation sources operating in the visible wavelength range.

The taking into account of an image in reflection and / or of a transmission image, acquired by means of a radiation source in the wavelength range of the visible and / or with other wavelengths such as that wavelengths in the infrared range depends in particular on the absorption and reflection properties of the thin-film stack considered, and characteristics representative of the defect considered for implementing the invention. This is further illustrated with reference to the given examples.

It is assumed that the location device 6 obtains from the optical control system 4 at least one greyscale-encoded IM image of the detected fault on the thin-layer stack deposited on the glass substrate 3 (step E10). This IM image or images are received by the location device 6 via its communication module 23 and its obtaining module 6A.

The locating device 6, via its determination module 6B, then determines from the IM image (s), a signature SIG (DEF) of the defect (step E20) .By signature of the defect, is understood to mean of the invention one or more featuresrepresentatives of the defect and which will allow to locate its origin in the filing line 2 approximating reference signatures associated with each of the compartments 15 of the deposit line. These characteristics may depend in particular on the nature of the IM images exploited by the localization device 6 and the information on the defect contained in these images, the reflection and absorption properties of the stack of thin layers deposited on the substrate, the shape of the defect, etc.

To better illustrate this step, we first consider the stack of thin layers of Example 1 described above and comprising a thin layer of silver framed two modules Ml, M2 each formed by a plurality of thin layers based dielectric materials (eg layers of oxide, nitride, oxynitride). It is furthermore assumed that the optical control system 4 provides a single image in reflection IM = IMR coded in gray levels, acquired by means of the light source 18-2 operating in the range of visible wavelengths.

In the embodiment described here, the determination module 6B extracts from this image in IMR reflection various characteristics relating to the light reflection coefficient of the defect. For this purpose, it applies here, on the image in IMR reflection, a method of erosionsuccessive allowing it to extract during several iterations different values of the coefficient of reflection of the defect.

FIG. 6 schematically illustrates the application of this method to a defect DEFrepresented by an IMR reflection image. This method consists, from the reflection IMR defect image, to "erode" or "eliminate" successively the defect on its contour (ie at each iteration iter eliminates a small thickness of the defect contour, this thickness being not necessarily uniform from one iteration to another) and calculating the average of gray levels of pixels on the surface of the defect eliminated, the gray level of a pixel reflecting the light reflection coefficient of the defect at the location represented by the pixel. The area remaining after each erosion is shown for illustrative purposes in Figure 6 in the frames associated with each iteration (iter = 1, iter = 2, ..., iter = 7). The number of iterations and the erosion step applied to each iteration (for example 1 pixel) used to implement this method of erosionsuccessive are chosen in coherence with the shape of the defect in order to allow the extraction of relevant information on this defect. .

In the example illustrated in Figure 6, the defect is represented by a task surrounded by a clear ring. At the first iteration (iter = 1), the eroded and considered surface for calculating the average pixel greyscale voluntarily includes only a "black" part of the image's bottom located outside the defect and surrounding it. This results in a relatively low average of gray levels (because only integrating very low gray levels representative of the black surface surrounding the defect). At the second iteration (iter = 2), the eroded part of the defect corresponds to the ring clearing the defect. This clear ring corresponding to high gray levels, resulting in a coefficient of maximum light reflection.

Then the erosion is continued and the mean of the luminous reflection coefficients decreases to stabilize in the example represented from the fourth iteration (iter = 4). "Stabilized" value of the average over the last iterations gives the value of the light reflection coefficient of the defect, denoted RL (DEF). The presence of a maximum, at the second iteration, denoted RLmax (DEF) and the peak pace of the average of the gray levels illustrates the presence of a clear ring around the defect. The erosion method thus described makes it easy to identify the presence of a clear ring around the defect. It is noted that in the absence of a clear ring, the values RLmax (DEF) and RL (DEF) are substantially the same.

In the first embodiment described here, the value of the reflection coefficient RL (DEF) is a characteristic of the defect used by the determination module 6B to determine its signature. The light reflection coefficient RL (DEF) considered in the signature is preferably normalized, for example with respect to the light reflection of the coated substrate 5 resulting from the deposition line 2 (final product) or with respect to the luminous reflection of the bare glass (average contrast ). Such a standardization advantageously makes it possible to overcome adjustment fluctuations of the various cameras of the optical control system. It is optional, however. In the remainder of the description, it will thus be possible to consider indifferently normalized or non-standardized reflection and / or transmission coefficients, the invention applying indifferently in these two cases.

The determination module 6B also adds to the signature of the defect an indicator of the presence or not of a clear ring around the detected defect as indicated previously from the value RLmax (DEF).

Alternatively, other techniques than a successive erosion method can be used by the determination module 6B to determine the signature of the defect and to detect whether or not a clear ring is around it. For example, the determination module 6Bcan determine a light intensity profile of the defect by evaluating the reflection coefficient at different points of a diagonal of the defect (eg the largest diagonal or the smallest diagonal). The analysis of the different values thus obtained from the intensity of the light intensity of the defect enables the determination module 6B to detect the presence of a clear ring around the defect and to evaluate the average value of the light reflection coefficient of the defect.

According to yet another variant, intensity profiles can be determined by the determination module 6B over a plurality of fault radii, and then averaged to extract the signature of the defect.

Then the signature of the defect thus obtained is compared by the identification module 6C of the localization device 6 with several reference signatures (step E30) to identify the compartment or compartments likely to be at the origin of the defect.

In the embodiment described here, the identification module 6C uses a reference sign SIGref (i) for each compartment 15-i, i = 1, ..., N distinct from the deposition line 2. This reference signature comprises the same characteristics (ie the same types of characteristics) as the signature of the defect. The reference signature may comprise a single value (i.e. unique point) for each characteristic or on the contrary comprise a range of values for each characteristic, or only the bounds or some significant values of such a range of values.

Alternatively, a plurality of reference signatures may be associated with a same compartment.

In the first embodiment described here, the reference signatures are previouslygenerated (step E00) and stored for example in the non-volatile memory 22 of the location device 6. Alternatively, they can be stored on a storage space and be obtained on request for example via the communication module 23 of the localization device 6.

The reference signatures are generated, for example, experimentally, from a noise made on each of compartments 15-i, i = 1,..., N of the deposit line 2. This procedure consists in typing the walls and the elements of each of the compartments and to recover the debris resulting from this fuss. The observation by means of the optical control system 4 of the debris thus obtained for a compartment and the analysis of the images resulting from this observation makes it possible to estimate an average reference signature associated with the compartment having the same characteristics as those extracted for the defect ( luminous reflection coefficient, presence or absence of a clear ring in the example envisaged here). In other words, at the end of step E00, the locating device 6 has in memory memory reference signatures for each of the compartments 15-i of the deposit line, these reference signatures giving for each compartment 15-i a reflection coefficient average luminous said reference (if appropriate, normalized), averaged on several debris resulting from the noise made on the compartment 15-i, and indicator reflecting on average, the presence or not of a ring clear on the periphery of the defects originating from this compartment.

Alternatively, the indicator of presence or absence of a clear ring included in the reference signature of each compartment 15-i can be determined according to the position ofcompartment 15-i relative to the compartments involved in the deposition of the layer of Indeed, the inventors have judiciously established a rapprochement between this position and the presence or not of a clear ring around the defect: because of the high emissivity of the silver layer, the presence of a clear ring around the defect on a reflection image of a defectassociating a stack of thin layers comprising a single layer of silver translates the probable appearance of this defect before the deposition of the silver layer (ie in a compartmentparticipant in the deposit of the module Ml). Conversely, the absence of such a ring reflects the probable appearance of this defect after the deposition of the silver layer (i.e. in a compartment participating in the deposition of the module M2). When determining the reference signatures, it is also possible to directly affect the compartments participating in the deposition of the layers forming the module Ml, an indicator reflecting the presence of a clear ring, and to the compartments participating in the deposition of the layers forming the module M2 an indicator reflecting the no clear ring

It is noted that the reference signatures thus generated depend on the parameterization of the deposition line 2 and in particular the sputtering processes implemented by the cathodes of the different compartments 15-i. Thus, they depend in particular on the power applied to each cathode, the deposit thickness to be deposited by each cathode, the gas mixture in the compartment, etc. The reference signatures can be determined experimentally for each condition (each parameterization) of deposition, leading noisy experiments under these different conditions.

In a variant, it is possible to envisage the use of a calculation model allowing the switch from one parameterization to another.

Such a calculation model is for example the following, when the reference signatures correspond to intervals of values. Its application to a single value is immediate for the person skilled in the art and is not described here.

We denote by e (k, x, y) the contribution of a cathode k depositing a material x on the substrate in a module y. This contribution is assumed here proportional to the electric power injected into this cathode, denoted P (k, x, y). We also note E (x, y) the total thickness of the layer of material x in the module y.

To calculate the range of signatures corresponding to the cathode k, we can calculate the cumulative limits limiting the contribution of the cathode k (i.e. corresponding to the regions of the reference signature of the cathode k), ie: for the lower bound:

- for the upper limit:

where £ (%; y) represents the sum of the thicknesses of the layers deposited before the layer of material x.

If F denotes the function which makes it possible, from the cumulative thickness, to deduce the reference signal from a compartment (F is for example a vector function providing the limits of the curve portions defining the reference signatures, or a function calculating the averages on the curve portions), then the reference signatures corresponding to the new parameterization are obtained by applying the function F to the lower limits e (k1, x, y) and higher e (k, x, y) given above.

In an alternative embodiment, the reference signatures are generated by simulation or by calculation.

FIG. 7 illustrates an example of reference reflection coefficients Rref measured on the side of the stack and obtained by fussing, and corresponding to the deposition of a thin-layer stack according to example 1 in which a plurality of cathodes identified as abaxes (K cathodes) participate; , K3, K4, etc.), these cathodes being included in as many compartments (ie there is an unequivocal relationship between the cathode indicated on the abscissa and a compartment of the deposit line).

In this example, the reference reflection coefficients Rref were generated from images acquired by four separate cameras of the optical control system 4 placed to monitor different areas of the coated substrate. During the generation of the reference reflection coefficients Rref, it is possible either to estimate a reference coefficient per camera or zone of the coated substrate, or to estimate a hybrid reference coefficient from the reference coefficients obtained by camera, for example resulting from the average of the reference coefficients obtained by camera.

In the example illustrated in FIG. 7, a single reference reflection coefficient value Rref for each cathode is determined. As a variant, it is possible to determine for each chase not a single reference reflection coefficient value but an interval comprising several possible values of the reference reflection coefficient, or in another still variable the limits of an interval of possible values of the reflection coefficient. reference.

It is noted that in FIG. 7, the cathodes K9, K9B and K10A correspond to the cathodes involved in the deposition of the silver layer. These cathodes are known not to generate (or generate little) debris. In other words, it is unlikely that a defect affecting the thin film stack originates from a compartment comprising one of these cathodes. Note also in the example illustrated in Figure 7 that some cathodes (or equivalent compartments) are associated with the same value of reflection coefficient (eg Cathodes K4 and K23).

The reference reflection coefficients Rref may alternatively be represented not according to the compartments with which they are associated but of a deposition thickness in the stack. In this case, a correspondence between the thickness of the deposit and the compartment participating in the deposit corresponding to this thickness is established and stored at the location device 6 for example in its non-volatile memory 22. The same compartment can be associated with one or more several deposit thicknesses, in other words, one or more reference signatures.

In the first embodiment described here, the comparison of the default signature with the reference signatures associated with the different compartments of the debit line 2 is carried out in two stages by the identification module 6C.

More specifically, in a first step, the identification module 6C determines from the signature SIG (DEF) of the defect if it has a clear ring around its periphery.

For the reasons mentioned above, it is assumed here that the reference signatures associated with the compartments of the deposition line 2 located before the compartments containing the deposition cathodes of the silver layer (ie deposition compartments of the thin layers forming the module Ml containing the Cathodes K1, K3, K4, K5, K5B, K6, K7 and K8 in the example illustrated in FIG. 7) include among their characteristics an indicator reflecting the presence of a clear ring around the periphery of a defect originating from these compartments.

If the identification module 6C detects from the defect signature SIG (DEF) the presence of a clear ring around the periphery of the defect, the identification module 6C therefore associates the defect with one of the compartments participating in the training. of the module Ml (and corresponding to one of the cathodes K1, K3, K4, K5, K5B, K6, K7 and K8 in the example illustrated in FIG. 7).

If, on the other hand, the signature of the defect SIG (DEF) indicates the absence of a clear ring on the edge of the defect, the identification module 6C associates the defect with one of the compartments of the deposit line located after the compartments. containing the cathodes participating in the deposition of the silver layer, that is to say to one of the compartments participating in the deposition of the module M2 andcontaining a cathode among the cathodes K11, K12, K13, K19, K21, K27B, K28 , K30.

Then, in a second step, the identification module 6C compares the reflection coefficient of the defect DEF contained in its signature SIG (DEF) with the reference coefficients of reference Rref reported in the reference signatures of the compartments previously selected (or the intervals of values or the limits of the intervals defining the reference reflection coefficients). It then selects the compartment whose reference reflection coefficient Rref corresponds to the reflection coefficient (or the interval or the terminals) reported in the signature of the defect, that is to say the one whose reference signature corresponds to the signature of the defect. By "corresponds to" is meant here the compartment whose signature reference is closest to the signature of the defect, for example in the sense of a predetermined distance. For this purpose, in the example illustrated in FIG. 7, of a single value of the reference reflection coefficient included in each reference signature, the identification module 6C selects the compartment from among all the compartments 15-i, i = 1, .., N of the deposition line 2minimitating a predefined distance, such as an Euclidean distance (also called "L2 type distance") or an absolute distance (also called "L1 distance"). This resistance is calculated between the reflection coefficient contained in the signature of the defect SIG (DEF) and the reference reflection coefficient Rref contained in the signature of the reference SIG associated with the compartment in question.

Note that when the reference signature comprises not a single values but an interval of values or boundaries of such an interval, the identification module 6C can equivalently evaluate a distance from the signature of the defect to that interval or cesbornes. and select the compartment that minimizes this distance.

The compartment thus selected is identified by the identification module 6C as the compartment causing the fault DEF affecting the coated substrate 5 (step E40).

Following this identification, a maintenance operation may be undertaken on the identified compartment. This operation may include cleaning this compartment, for example by tapping on these walls to detach the debris that is deposited there.

It is noted that the analysis of the digital images of the defect provided by the optical control system 4 to the locating device 6 can make it possible to identify other useful information on the defect.

For example, during the step E20 of determining the signature of the defect, the determination module 6B can determine a gradient of variation of the reflection coefficient of the defect. This gradient can be easily determined from the erosion curve obtained in FIG. implementing the successive erosion method or values of reflection coefficients obtained when caséchéant on a diagonal or on radii of the defect. From the knowledge of this gradient and its variations, the determination module 6B can deduce information on the form of the defect. It can notammentidentify if the defect is flat (or relatively flat) when the gradient has a maximum on a low number of pixels (typically on a pixel), or if it has a growth in three dimensions when the gradient is maintained at a quasi-constant value. constant over a larger number of pixels.

The knowledge of this information on the form of the defect facilitates the identification even of a compartment of the element at the origin of the debris that created the defect, and optimizes the maintenance operation performed on the compartment. There are indeed different protective shields in each compartment around the cathode, as shown in Figure 8 by reference 25. Depending on whether the debris comes from rough areas of these shields or flat areas, they do not have the same shape and engender on the substrate different defect forms. By way of illustration, a debris originating from a shield without a grid placed at the level of the ceiling of the compartment will generate a defect of planar shape. On the other hand, debris from a horizontal shield on which there is a metal grid with growths and located in the bottom of the compartment will cause a defect in three dimensions.

It is noted that although it is described from an image in reflection coded in degrees gray, acquired from a radiation source emitting radiation in the field of visible wavelengths, the first embodiment which comes from may also be presented using a grayscale reflection image acquired from a radiation source operating in another wavelength domain (eg infrared) or a transmission image encoded in gray levels, obtained from a source of radiation emitting radiation in a range of wavelengths adapted reflection and absorption properties of the thin film stack considered.

In the example illustrated in FIG. 7, for a stack of thin layers according to example 1, it suffices for the localization device 6 to analyze the image in IMR reflection of the fault to identify which compartment is the source of this the uncertainty of the location of the compartment between the series of compartments participating in the formation of the module M1 and the series of compartments participating in the formation of the module M2 can be easily removed as previously described by the detection of the presence or absence of a ring clear on the blemished defect.

Although described with reference to a stack according to Example 1, this first embodiment is applied to other thin film stackings forming an interference system, and in particular to the stack of Example 3 as well as to FIG. other stacks comprising a single functional layer.

It can also be applied when the stack of thin layers comprises several functional layers, for example two thin layers of silver as in Example 2 introduced previously.

However, for such a stack, it may be necessary to take into account several images of the defect (eg a reflection image and a transmission image) if it is desirable to identify more precisely the compartment at the origin of the defect. affecting the stacking of thin layers.

FIG. 9A illustrates an example of reference reflection coefficients Rref of the two compartments participating in the deposition of a stack of thin layers according to example 2.The reference reflection coefficients Rref are measured on the stacking side and are normalized with respect to the reflection coefficient of the coated glass substrate 5 obtained at the outlet of the deposition line 2. They are represented as a function of the deposition thickness on the substrate 3, expressed in nanometers (nm). As mentioned above, a correspondence between the thickness of the deposit and the compartment participating in the deposition of a thin layer associated withthis thickness may be pre-established and stored in the nonvolatile memory 22 of the delocalization device 6 to enable it to identify the compartment at the same time. origin of the defect.

The reference reflection coefficients Rref of the compartments participating in deposition of the silver layers, which generate little debris, are identified in FIG. 9A by the references Agi (first silver layer corresponding to a thickness of deposition on the substrate varying between 41 nm and 48 nm) and Ag2 (second silver layer corresponding to a thickness of deposits on the substrate varying between 123 nm and 143 nm) respectively.

As it appears in this figure, a value of reflection coefficient equal to 0.1 can correspond to different (ie three) different deposit thicknesses and therefore to differentcomparisons of the deposit line 2. It follows that if the implementation of the The invention for stacking according to Example 2 is limited to considering a signature of the defect consisting of a single characteristic corresponding to the reflection coefficient of the defect, the identification step E40 can lead, depending on the value of this reflection coefficient, identifying a plurality of compartments. In the example illustrated in FIG. 9A, the identification leads to three compartments. It is noted that this number, although not unitary, corresponds to a relatively small number of compartments with respect to all the compartments belonging to the deposit line, which makes it possible to simplify the maintenance implemented on the deposit line. .

Taking into account the presence of a clear ring around the defect as described previously in the signature of the defect can make it possible to distinguish a defect originating from a compartment which participated in the deposition of the thin layers of the module Ml of a compartment having participated in the filing of the thin layers of the M2 module. However, this consideration may be insufficient to distinguish two compartments that participated in the deposition of the thin layers within the same module (in this case M2 in the example illustrated in FIG. 9A).

In order to overcome this insufficiency, it is possible to consider for example 2, in addition to the image in IMR reflection, an image in IMT transmission acquired by means of a radiation source emitting in the wavelength domain of the infrared (for example at a wavelength of 850 nm). This radiation source can be placed in the optical control system 4 instead of for example the light source 18-3, or belong to another optical control system placed at the output of the deposition line 2. It can be a different broadband source covering at least a part of the infrared wave length range, the optical control system then being preferably provided with a spectrally resolved camera for acquiring an infrared image, or a spectrally resolved radiation source, the system The optical control device can therefore be provided with a broadband camera or a spectrally resolved camera in the infrared range.

The determination module 6B can therefore extract, from this IMT transmittance image in the infrared domain, a transmission coefficient T (DEF) of the defect, in a manner similar to or identical to that described above for extracting the deflection coefficient of the defect from the IMR reflection image. This transmission factor, possibly normalized, is used as a characteristic of the signature of the defect in addition to the normalized reflection coefficient extracted from the IMR image. The signature of the defect SIG (DEF) is thus formed on the one hand, of the defect reflection coefficient determined from the IMR image and, on the other hand, of the transmission coefficient of the defect determined from the IMT image. It is not envisaged here in the signature of the defect of presence indicator of a clear ring.

FIG. 9B illustrates the reference transmission coefficients Tref of the compartments of FIG. 9A participating in the deposition of the thin-film stack according to example 2. These reference transmission coefficients Tref can be obtained as described previously for the reflection coefficients of FIG. Referring to FIG. 9B, reference transmission coefficients Tref are normalized with respect to the transmission coefficient of the coated glass substrate obtained at the output of the line. 2. They are represented as a function of the thickness of deposition on the glass substrate 3, expressed in nanometers (nm). The reference transmission coefficients Tref of the compartments participating in deposition of the silver layers are identified in FIG. 9B by the references Agi (first silver layer corresponding to a substrate deposition thickness varying between 41 nm and 48 nm) and Ag2 ( second silver plating corresponding to a deposition thickness on the substrate varying between 123 nm and 143 nm) respectively.

As it appears in this figure, a Trefential reference transmission coefficient at 0.71 leads to two possible deposit thicknesses. In other words, taking into account a transmission coefficient in the signature of the defect could lead, depending on the value of this coefficient, to identify two compartments potentially at the origin of the defect.

In order to have a more precise estimate of the location of the compartment causing the defect, the identification module 6C takes into account in the signature of the defect both the reflection coefficient of the defect and the transmission coefficient of the defect determined by the module. determination 6B. Each reference signature is then similarly composed of a reference reflection coefficient value Rref and a reference transmission coefficient value Tref.

FIG. 9C illustrates, for different deposit thicknesses (the same as those considered in FIGS. 9A and 9B), the reference reflection coefficient Rref (in the abscissa) under the reference transmission coefficient Tref belonging to the same signature of reference. It appears in this FIG. 9C that all the reference signatures corresponding to the different deposit thicknesses envisaged are distinct.

In other words, the comparison of the signature of the defect SIG (DEF), consisting of the coefficient of reflection of the defect and the coefficient of transmission of the defect, with the reference signatures associated with the compartments enables the identification module 6C to identify a unique compartment at the end of the period. origin of the defect. This compartment is associated with the signature of reference closest to the signature of the defect, that is to say, minimizing a predefined distance, such as for example the Euclidean distance between the two vectors corresponding to thecompared signatures, each vector having for components a reflection coefficient and a transmission coefficient.

The taking into account of several characteristics in the signature of the defect and in the reference signatures is however not mandatory in the case of stacking such as the stacking of example 2. In fact, the inventors have found that by using a transmission image acquired at a wavelength judiciously chosen in the infrared range with respect to the absorption and reflection properties of the stack, it is possible from a signature comprising a single characteristic defined from a transmission coefficient of the defect extracted from this image, to precisely identify the compartment at the origin of the defect.

FIG. 10 illustrates an example of standardized reference transmission coefficients Tref normalized of compartments participating in the deposition of a stack of thin layers according to example 2 and extracted from images in transmission of defects resulting from a noise on the compartments. In the example of FIG. 10, the transmission images considered have been acquired by means of a radiation source operating at a wavelength of 1050 nm, thus belonging to the infrared range.

It appears in this figure that these reference transmission coefficients Tref are all distinct as a function of the thickness of the deposit. In other words, the comparison of a fault signature consisting of a single transmission coefficient extracted from a transmission image acquired at the same wavelength of 1050 nm makes it possible to uniquely identify the compartment at the origin of the fault.

In the first embodiment that has just been described, the optical control system 4 supplies the location device 6 with grayscale-coded digital images, and the location device 6 uses various information contained in these images to determine the origin. However, the invention applies to other types of images, as now described in the second and third embodiments.

FIG. 11 represents the main steps of the localization method according to the invention as implemented by the localization device 6 in the second embodiment.

In this second embodiment of the invention, the optical control system 4is equipped with color cameras and provides trichromatic digital images encoded in red green and blue (RGB or RGB for Red Green Blue). By RGB images is meant here preferably images coded according to the sRGB standard. In a manner known per se, according to the RGB encoding, each point or pixel of the image is encoded by means of three magnitudes, respectively indicating the intensity of each of the primary colors red, green and blue for this point.

In the second embodiment, we are interested in RGB-encoded defect images acquired by the control system 4 by activating the RBF-type light source 18-2, which images reflect a defect reflection coefficient detected on the coated substrate. 5, taken from the stack. For the sake of simplification, the following description is limited to a single IMR reflection image of the defect. However, a similar approach can be applied from a transmission image.

The locating device 6 obtains from the optical control system 4 the RGB-encoded IMR reflection image of the detected defect on the stack of thin layers deposited on the glass substrate 3 (step F10). This IMR image is received by the location device 6 through its communication module 23 and its obtaining module 6A.

In the second embodiment described here, before extracting the reference signature of the defect DEF from the received IMR image, the locating device 6 proceeds to convert this image into a different color space than the RGB space (step F20 ). More specifically, it performs a conversion of the IMR image in the color space a * b * L * familiar to the skilled person. In this color space, L * represents clarity, and the components a * and b * characterize the deviation of the color of the point considered from that of a surface of similar clarity. Note that in this color space, L * corresponds more or less to the level of light reflection of the defect.

The image conversion is carried out here in two stages by the localization device 6: in a first step, it proceeds to the conversion of the IMR image encoded in RGB into a esp, Υ, Ζ spacecarecorimetric space, for example as described in FIG. detail on the websitehttD: //www.brucelindbloom.com/index.html? Waters.html. Then it proceeds to a conversion of the image obtained from the color space (Χ, Υ, Ζ) to the color space (L * a * b *), as indicated in detail on the aforementioned website or on the site httDs: / /en.wikiDedia.orq/wiki/CIE L * a * b * CIE XYZ conversions to CIE L.2Aa.2Ab.2A.

Following this conversion, the locating device 6, via its determination module 6B, determines from the converted image (denoted IMR ') the signature SIG (DEF) of the defect (step F30). For this purpose, it extracts from the converted image IMR 'the coordinates according to the components a * and b * of a bottom surface of the defect. It can use for this purpose a successive erosion method as described above. These are aDEF * and bDEF *.

Then the signature of the defect thus obtained comprising the components aDEF * and bDEF * is compared by the identification module 6C of the localization device 6 with several reference signatures (step F40), each reference signature SIGref (i) being based on the same characteristics the signature of the defect and being associated with a compartment 15-i, i = 1, ..., N of the deposit line 2, or at a deposit thickness which is itself connected to a compartment.

The reference signatures may be previously generated and stored, for example, in the non-volatile memory 22 of the localization device 6. In a variant, they may be stored on a remote storage space and be obtained on request, for example via the communication module 23 of the communication device. location 6.

As in the first embodiment, the reference signatures can be generated from a noisy experiment, performed beforehand on the compartments of the deposition line 2 (step F00). The observation with the optical control system 4 of the debrisgenerates in each compartment and in particular the images in RGB reflection of these debris, makes it possible to obtain on average coordinates aref * (i) and brief * (i) of reference for each compartment 15-i, i = 1, ..., N, or for different deposit thicknesses associated with the different compartments. It will be noted, as mentioned previously for the first embodiment, that the correspondences between the reference signatures and the different compartments depend on the parameterization of the deposition line 2 and in particular the cathodic sputtering parameters in the different compartments 15-i.

FIG. 12 illustrates an example of reference coordinates (aref *, brief *) obtained for different thicknesses of thin layers deposited in the different compartments 15-i, i = 1,..., N of the deposition line 2 and leading to a stack of thin layers according to the example3 previously described. In this figure, the x-axis represents the aref * component and the y-axis the short component *. Each thickness corresponding to the deposit made by a cathode of one of the compartments is illustrated by a symbol, the shape of the symbol depending on the material deposited by this cathode (eg rhombus for the thicknesses corresponding to the deposition of the TiO2 layer, square for the corresponding thicknesses deposition of the SiO2 layer (L2), etc.). For the sake of readability, the deposit thicknesses corresponding to each pointreporté in the figure have been omitted.

In addition, for the sake of simplification, it is assumed here that each reference signature associated with a thickness is composed of a single value of the components aref * and brief * .However, as mentioned above, it is possible to envisage as a variant that each thickness is associated with an interval of values aref * and at a range of short values *, or at the boundaries defining such intervals.

The comparison of the signature of the defect SIG (DEF) with the reference signatures SIGref (i), i = 1,..., N is carried out very simply during step F40 by searching for the thickness of stacking whose signature reference is closest to the signature of the defect. This thickness is then connected as previously indicated to the compartment or the cathode corresponding to a deposit at this thickness. In other words, the identification module 6C searches on the reference curve illustrated in FIG. 12 for the point (aref *, brief *) closest to the point (aDEF *, bDEF *) (illustrated by a cross in FIG. 12). , and deduces the deposit thickness and the associated compartment.

Alternatively, when the reference signatures comprise an interval of values for aref * and brief * or the boundaries of such intervals, the identification module 6C looks for the nearest reference signal in the sense of a predefined distance from the point (aDEF * , bDEF *) and deduces the deposit thickness and the associated compartment. No limitation is attached to the resistance considered: it may be in particular the distance from the normal to the curve defined by the reference intervals considered for aref * and brief * passing through the point (aDEF *, bDEF *).

The compartment thus determined is identified by the identification module 6C of the location device 6 as the compartment likely to be at the origin of the defect DEFaffectant the coated substrate 5 (step F50).

Following this identification, a maintenance operation can be undertaken on the compartment thus identified.

Note that this second embodiment can also be applied in other stack configurations such as in a configuration similar to Example 2 and which comprises two layers of silver.

In such a configuration, in case of ambiguity existing on the precise localization of the compartment, the determination module 6B of the signature of the defect can add to the default signatured the component LDEF * of the bottom surface of the defect, and generate the equivalent component Lref * in the reference signatures of the sub-funds. A three-dimensional curve is then obtained illustrating the reference signatures of compartments 15-i, i = 1,..., N (or deposit thicknesses corresponding to these compartments), as illustrated in FIG. 13 for another example of FIG. stack of thin layers with two layers of silver.

During the comparison step F40, the identification module 6C looks for the reference signatured (point (Lref *, aref *, brief *)) on this curve closest to the signature of the defect (LDEF *, aDEF *, bDEF *) (represented by a cross in FIG. 13). The identification module 6C then identifies the compartment corresponding to the reference signature thus determined as the compartment causing the defect DEF.

FIG. 14 represents the main steps of the localization method according to the invention as it is implemented by the localization device 6 in a third embodiment.

In this third embodiment of the invention, the optical control system 4 is equipped with hyperspectral cameras and provides hyperspectral digital images of the defect detected on the coated substrate 5. These hyperspectral images are remarkable in that they associate with each pixel of the image a spectrum over a range of lengths of data (spectrum in reflection or in transmission depending on the configuration of the radiation sources considered). Each spectrum is defined as a vector of K components Sp (Xk), k = l, ..., K where K denotes an integer greater than 1, each component Sp (Xk) denoting the value of the reflection (or reflectance) or of the transmission of the fault at the wavelength Xk.

The locating device 6 thus obtains from the optical control system 4 a hyperspectral image in IMR reflection of the defect detected on the stack of thin layers deposited on the glass substrate 3 (step G10). This IMR image is received by the location device 6 through its communication module 23 and its obtaining module 6A.

The locating device 6, via its determination module 6B, then determines from the hyperspectral image IMR the signature SIG (DEF) of the defect (step G20). For this purpose, it extracts from the IMR image the spectrum in reflection of the bottom surface of the defect, for example by means of a successive erosion method as presented previously in the first embodiment of the invention. and applied to the reflection spectrum of each pixel (a reflection coefficient value is extracted for each wavelength of the reflection spectrum associated with each pixel instead of a single reflection coefficient value per pixel as in the first embodiment ).

The signature SIG (DEF) determined by the determination module 6B is therefore composed of K spectral components SpDEF (Xk), k = 1,..., K thus extracted.

Then the signature of the defect SIG (DEF) thus obtained is compared by the identification unit 6C of the location device 6 to several reference signatures (step G30), each reference signature SIGref (i) being based on the same characteristics as the signature of the defect and being associated with a compartment 15-i, i = 1, ..., N of the deposit line 2.

As in the first two embodiments, the reference signatures can be generated previously and stored for example in the non-volatile memory 22 of the location device 6. Alternatively, they can be stored on a storage space and be obtained on request for example via the communication module 23 of the localization device 6.

As in the first and second embodiments described, the reference signatures can be generated from a noisy experiment, performed beforehand on the compartments of the deposition line 2 (step G00). The observation by means of the optical control system 4 of the debris generated by each compartment and in particular the hyperspectral images in reflection of these debris, makes it possible to obtain Spref reference spectra for each compartment 15-i, i = 1, ... , N, or for different deposit thicknesses themselves associated with compartments 15-i, i = 1, ..., N.

Moreover, as mentioned above for the first and second embodiments, the correspondences between the reference signatures and the different compartments depend on the parameterization of the deposition line 2 and in particular the cathodic sputtering processes in the different compartments 15-i. .

FIG. 15 illustrates an example of reference spectra obtained for different thicknesses of thin layers deposited by the compartments 15-i, i = 1,..., N of the deposition line 2, for a stack of thin layers according to example 3 previously described. In thisfigure, the different spectra are given as a function of the thickness of the deposit on the substrate. In other words, at the same compartment 15-i, corresponds a set of reference spectraSpref (j), j = Jl (i) ,. .J2 (i), where Jl (i) and J2 (i) denote two integers depending on the index i, this set of spectra being easily identifiable from the thickness of the deposit of which this compartment 15-i is loaded.

Each reference signature SIGref (i) associated with a compartment 15-i therefore comprises a plurality of spectra (Spref (Jl (i)) ... Spref (J2 (i))), the number of spectra J2 (i) -Jl (i) + 1 may vary from one compartment to another depending on the thickness of the thin layer deposited by the compartment) and the step of discretization envisaged, and each spectrum Spref (j), j = Jl (i), ..., J2 (i) being defined by a vector of K components.

The comparison of the signature of the defect SIG (DEF) with the reference signatures SIGref (i), i = 1,..., N is carried out during the step G30 by the identification module 6C encalculating a distance (by example defined according to a standard of type L2) between the signature of the defect SIG (DEF) and each reference signature SIGref (i). This distance is defined, in the third embodiment described here, by:

where j = l, ..., Jl (1), J1 (2), ..., J2 (N).

As a variant, each reference signature associated with a compartment 15-i can comprise only two limit spectra flanking on each side the surface portion defining the spectra associated with this compartment. In this case, the identification module 6C can proceed in two stages by first identifying a first limit spectrum among the reference signatures corresponding to the spectrum closest to the signature of the defect. Since two neighboring compartments share the same limit spectrum, it is then sufficient for the identification module 6C to identify between these two adjacent compartments, which corresponds to the signature of the defect.

Alternatively, other distances may be used to compare the default signature with the reference signatures, for example a distance based on an L2 standard such as that indicated above but which includes the application of a different weighting factor for wavelengths (indexed by k) belonging to the ultraviolet domain.

The identification module 6C identifies the compartment associated with the index j minimizing this distance (step G40) as being the compartment of the deposition line 2 at the origin of the defect affecting the coated substrate 5.

Following this identification, a maintenance operation can be undertaken on the compartment thus identified.

Note that this third embodiment can be applied identically in other stacking configurations such as in a configuration similar to Example 2 and which comprises two layers of silver.

In the first three embodiments described above, the delocalization device 6 determines a signature of the defect from one or more characteristics representative of reflection and / or absorption properties of the defect, namely in particular a coefficient of reflection extracted from an image in reflection, a transmission coefficient extracted from a transmission image, and compares this default signature with reference signatures associated with each compartment of the deposition line 2, obtained for example via an experience of noise. According to the absorption and / or reflection properties of the coated substrate 5 and the type of images provided by the optical control system (images in reflection, in transmission, acquired by means of radiation sources emitting in the range of lengths d visible wave or infrared, etc.), a single characteristic may be sufficient in the signature to identify a single compartment likely to be at the origin of the defect or on the contrary several features may be necessary to identify this single compartment. However, in such a context, taking into account a single well-chosen characteristic of the

The default already allows for a drastic reduction in the number of compartments likely to be at the origin of the default, compared to the total number of compartments included in the deposit line.

These three embodiments are somewhat similar to considering a decisional arbr to identify one or more compartments likely to be the cause ofdefault.

We will now describe a fourth embodiment in which the identification of the compartment causing the defect is implemented using a method of learning. In the fourth embodiment described here, this learning method makes it easy to take into account a plurality of decision trees.

FIG. 16 represents the main steps of the localization method implemented by the localization device 6, in this fourth embodiment.

It is assumed here, as in the first embodiment, that the optical control system 4 generates, for each defect detected at the output of the deposition line 2, digital images in reflection and / or transmission encoded in gray levels. For the sake of simplification, it is limited in the following description to an IMR reflection image encoded grayscale provided by the optical control system 4 for each detected fault. Note that this fourth embodiment also applies in the case where the optical control system 4 generates and provides RGB-encoded digital images or hyperspectral images.

The locating device 6 thus obtains from the optical control system 4 a gray level encoded IMR image of the detected defect on the stack of thin layers deposited on the glass substrate 3 (step H10). This IMR image is received by the location device 6 through its communication module 23 and its obtaining module 6A.

The locating device 6, via its determination module 6B, then determines from the IMR a signature SIG (DEF) of the defect (step H20) .Preferentially the signature of the defect SIG (DEF) comprises a or a plurality of light intensity characteristics of the defect and / or one or more characteristics relating to a defect shape.

More specifically, in the fourth embodiment described here, the default signature SIG (DEF) comprises: characteristics representative of a radial profile of light intensity in reflection (or transmission if the received image is a transmission image) of the defect. Such a profile may comprise a plurality of intensity values taken on the largest diagonal of the defect (for example 8 values, ie 8 characteristics); Characteristics representative of a gradient (slope) of a radial light intensity profile in reflection (or in transmission if the received image is a transmitting image) of the defect. Such characteristics are to be compared with the detection of the presence of a clear ring described in the first embodiment of the invention; - a representative characteristic of an area of the defect; A representative characteristic of a ratio of a perimeter of the defect on an area of the defect; A representative characteristic of a form factor of the defect; A representative characteristic of a mean light intensity in reflection (or transmission if the received image is a transmission image) of the defect; and - a characteristic representative of a reflection light intensity (or transmission when the received image is a transmission image) at the center of the defect. The extraction of these characteristics does not pose a problem in itself to the dummy man and is not described in detail here.

Of course, this list is not exhaustive and other characteristics can be extracted from the IMR image to represent the DEF defect.

Then the signature of the defect SIG (DEF) thus obtained is compared by the identification model 6C of the localization device 6 to reference signatures (step H30), each reference frame SIGref (i) being based on the same characteristics as the signature the defect and being associated with a compartment 15-i, i = 1, ..., N distinct from the deposit line 2.

In the fourth embodiment described here, the reference signatures have been previously degenerated from a so-called learning set consisting of a plurality of defect images detected at the output of the deposition line 2 acquired by the optical control system. 4 on several days of production. For each of these images, the signature of the defect is determined on the basis of the same characteristics chosen in step H20, and this reference (referred to as reference) is associated with the compartment at the origin of the defect or the corresponding deposit thickness. This compartment or this deposit thickness is determined experimentally, for example by making a noise on the compartments of the deposition line 2 for each detected defect, or by using a locating method according to the invention according to the first embodiment described. previously.

These reference signatures are then used, in the fourth embodiment described herein, to generate decision trees or decision trees (step H00). Knownly, such trees can be used as a predictive model to evaluate the value of a feature of a system since the observation of other characteristics of the same system. In other words, in the case envisaged here, it is a question of using these decision trees trained using the reference signatures of the compartments of the deposit line 2 to establish the original compartment of the defect from its signature. The decision trees are hereby created from the reference signatures extracted from the images of the training set and the compartment information associated with those signatures. These decision trees thus in themselves model the reference signatures of each of the compartments 15-i.

By using these decision trees and a learning algorithm based on such trees as for example a so-called randomdecision forest algorithm, the identification module 6C can automatically "classify" the fault signature. SIG (DEF), that is to say him to associate a reference signature or even a compartment of the deposit line 2 whose reference signature corresponds to the default signature. Such a learning algorithm is known per se and is not described further here. It is for example described in the document by T. Hastie et al. Entitled "The elements of statistical learning - Data Mining, Inference and Prediction", 2nd edition, Springer.

Alternatively, other algorithms for classifying elements between them (the elements here being signatures associated with compartments) can be considered, such as for example nearest neighbor algorithms, SVM (Support Vector Machine) or algorithms based on neural networks.

The identification module 6C identifies the compartment thus determined by the learning algorithm and corresponding to the signature of the defect as being the compartment with the origin of the fault (step H40).

Following this identification, a maintenance operation can be undertaken on the compartment thus identified.

The four previously described embodiments allow a quick and efficient identification of a compartment of a deposition line causing a defect affecting a stack of thin layers made by the deposition line. It should be noted that these embodiments have been described with reference to a deposition line implementing magnetron sputtering deposition. However, the invention applies to other deposition methods, likely to be affected by similar problems (regardless of the type of debris and defaults generated), such as other spraying methods such as beam spraying. Ion Beam Sputtering (IBS) or Ion Beam Assisted Deposition (IBAD), or vapor deposition, chemical vapor deposition (CVD) methods, plasma-enhanced chemical vapor deposition (PECVD for Plasma-Enhanced CVD), low-pressure vapor deposition (LPCVD for Low-Pressure CVD), etc.

Claims (8)

REVENDICAIIONS A method of locating, in a deposition line (2) comprising a succession of compartments (154), an origin of a defect affecting a stack of thin layers deposited on a substrate (3) in said compartments, wherein each thin layer of a material deposited in one or more successive compartments (154) of the deposition line and debris remaining on the surface of a thin layer deposited in a compartment act as masks for subsequent thin layer deposition and are at the origin of defects, the method comprising: - a step of obtaining (E10, F10, G10, H10) at least one image representing said defect acquired by at least one optical control system placed at the output of the deposition line; A determination step (E20, F30, G20, H20), from said at least one image, of a signature of the defect, said signature comprising at least one representative characteristic of the defect; An identification step (E40, F50, G4Q, H40) of at least one compartment of the debit line likely to be at the origin of the defect from the signature of the defect and using reference signatures associated with the compartment of the deposit line. 2. Method according to claim 1, in which the deposition of thin layers is carried out by the deposition line (2) on the substrate (3) by magnetic field assisted sputtering. 3. Method according to claim 1 or 2 wherein the identification stepcompared a step of comparing the signature of the defect with a plurality of signatures deferference each associated with the compartments of the deposit line, said at least onecompartiment identified as being susceptible to be at the origin of the defect being associated with a signature corresponding to the signature of the defect. 4. A method according to any one of claims 1 to 3 wherein: said at least one image comprises a grayscale-encoded reflection image (IMR), and the signature of the defect comprises a feature defined from a coefficient of reflection of the defect determined from the image in reflection; and / or - said at least one image comprises a transmission image (IMT) encoded in degrees, and the signature of the defect comprises a characteristic defined from a transmission coefficient of the defect determined from the transmission image. 5. The method of claim 4 wherein said coefficient is determined using a successive erosion method applied to the defect represented on the gray scale coded image. The method of claim 4 or 5 further comprising: - a step of determining a coefficient variation gradient; and a step of detecting a form of the defect from the determined variation gradient. 7. A method according to any one of claims 4 to 6, comprising a step of detecting a presence of a light ring on the periphery of the defect represented on said at least one image, the signature of the defect comprising a characteristic reflecting this presence. The method according to any one of claims 1 to 7 wherein said at least one image comprises two grayscale-coded images acquired by said at least one optical system using two radiation sources emitting in at least two wavelength ranges. partially distinct. The method according to any one of claims 1 to 3 wherein: said at least one image comprises a red, green and blue (RGB) coded reflection or transmission image, the method further comprises a conversion step (F10) RGB image in a L * a * b * chlorometric space; and the signature of the defect comprises components at a * and b * of a defect background surface determined from the converted image. 10. The method of claim 9 wherein the signature of the defect further comprises an L * component of the defect determined from the converted image. The method according to any one of claims 1 to 3 wherein said at least one image comprises a hyperspectral image and the signature of the defect comprises a spectrometer representing values of a reflection coefficient or a transmission coefficient of a surface of background of the defect according to a wavelength. The method of claim 11 wherein each reference signature associated with a deposition line compartment comprises a plurality of spectra corresponding to different thicknesses of the layer deposited in said compartment. 13. A method according to any one of claims 1 to 12 wherein reference tessignatures associated with the compartments of deposition sign are determinedexperimentally (EOO, FOO, GOO) from a fuss made on each of the compartments. 14. The method according to claim 1, wherein the identification step (H40) comprises an application (H30) of an automatic learning method to the signature of the defect, said learning method being based on on a driven model (H00) from the reference signatures associated with the compartments of the deposit line. 15. The method of claim 1 or 14 wherein the signature of the defect determined from said at least one image comprises at least one characteristic of light intensity of the defect and / or a characteristic relating to a form of the defect. The method of claim 15 wherein said at least one light intensity characteristic of the defect comprises: characteristics representative of a radiai light intensity profile of the defect; and / or - characteristics representative of a slope of a radial profile of light intensity of the defect; and / or - a characteristic representative of a mean light intensity of the defect; and / or - a characteristic representative of a luminous intensity at the center of the defect. 17. The method of claim 15 or 16 wherein said at least one feature relating to a form of the defect comprises: a characteristic representative of an area of the defect; and / or - a characteristic representative of a ratio of a defect perimeter to a defect area, and / or - a characteristic representative of a defect form factor. 18. The method as claimed in claim 1, in which the stack of thin layers forms an interference system. 19. A method according to any of claims 1 to 18 wherein the reference indicia associated with the deposition line compartments are dependent on a deposition of the deposition line to deposition of the thin films on the substrate. 20. A computer program comprising instructions for performing the steps of the locating method according to any one of claims 1 to 19 when said program is executed by a computer. 21. Computer-readable recording medium on which is recorded a computer program according to claim 20 comprising instructions for executing the steps of the locating method according to any one of claims 1 to 19. 22. Locating device (6) an origin of a defect affecting a stack of thin layers deposited on a substrate (3) in a plurality of compartments succeeding one another in a deposition line (2), where each thin layer of a material is deposited in a orseveral successive compartments (15-1) of the deposition line and debris remaining on the surface of a thin layer deposited in a compartment act as masks for the subsequent thin-film deposits and cause defects, œcomposition comprising A module for obtaining (6A) at least one image representing the defect acquired by at least one optical control system placed at the output of the line of deposit; A determination module (6B), from said at least one image, of a default signature, this signature comprising at least one characteristic representative of the defect; An identification module (6C) of at least one compartment of the deposit line capable of causing the defect from the signature of the defect and using reference signatures associated with the compartments of the deposit line . 23. System (1) comprising: - a deposition line (2) comprising a succession of compartments capable of depositing a stack of thin layers on a substrate (3), where each thin layer of a material is deposited in one or more compartments (15-i) successive deposition line and debris remaining on the surface of a thin layer deposited in a compartment act as masks for the deposition of subsequent thin layers and are at the origin ofdefects; At least one optical control system (4) placed at the output of the depot line configured to provide at least one image representing a defect affecting the stack of thin layers deposited on the substrate; and a locating device (6) according to claim 22, able to identify, among the compartmental succession of the deposit line, at least one compartment likely to be at the origin of the defect.